Files
Abstract
Vegetation mapping from high resolution image data is traditionally performed by extensive fieldwork and manual interpretation, which is time and labor consuming. Meanwhile, pixel-based image analysis is not suitable for high resolution imagery any more. This thesis applies an alternative approach, object-based image analysis (OBIA), to identify vegetation in the west of Thunderhead Mountain in the Great Smoky Mountains National Park. Definiens Professional is used for the task. After segmentation with an optimal scale parameter, various combinations of terrain characteristics, spectral and textural information are integrated to form the fuzzy classification rules. Four levels of vegetation class scheme are proposed and classified. The results indicate that OBIA is a promising approach to facilitating vegetation mapping and vegetation database construction. The incorporation of terrain characteristics, particularly elevation, improves classification accuracy. The use of fuzzy classification for multi-tiered vegetation mapping is potentially effective and promising. Classification accuracy is higher with coarser vegetation class schemes.